Inspiration
Note: I went through the problem statement and found out various facts about Blood Warriors. My highest respect to them for their hard work , dedication , time and effort they put into saving lives.
My journey began by analyzing the operational reality of community heroes like Blood Warriors. I am inspired by their life-saving mission but struck by a critical bottleneck: their staff spends the majority of their day making hundreds of calls and sending mass messages, a process that yields a response rate of only 2-5% . This operational inefficiency creates immense stress for patients and leads to volunteer burnout. I realized the most impactful problem wasn't a lack of donors, but the lack of an intelligent system to reach them effectively. This inspired me to think of a focused question: "How can we build an autonomous digital volunteer to handle the entire outreach workflow, freeing humans to do what they do best—connect and care?"
What we learned
The most important learning was the power of empathy and focused critical thinking. Thinking for a user persona is far more important than building the application itself. Instead of trying to build a large ecosystem/application , solving one critical operational bottleneck in current existing systems with a single, powerful tool will deliver exponential value. Automating the repetitive, time-consuming tasks of outreach doesn't replace the human element; it enhances it, allowing the Blood Warriors team to focus on patient care, community building, and strategic growth.
How I am going to build / Brief plan
Jeevan-Dhara , an Autonomous AI Agent that acts as a hyper-efficient digital operations manager for Blood Warriors. Given a single goal from a staff member (e.g., "Secure 2 units of B+ blood for Patient 123"), the agent independently plans and executes the entire outreach campaign. It uses a predictive model to identify the highest-potential donors, initiates contact, understands responses, and reports back only when the mission is complete or requires human intervention.
The architecture is designed for simplicity and immediate impact. The core of my project is built using Microsoft Autogen, allowing me to create conversational agents that can reason and execute tasks. This is supported by a lightweight Python/FastAPI backend and a simple React.js dashboard for admin supervision.
Challenges we ran into
The primary challenge is ensuring the safety and reliability of an autonomous agent handling sensitive, health-critical communications.
Agent Reliability & Safety
An agent making a mistake is not an option. The entire design is built around a "human-in-the-loop" safety protocol. The agent cannot act entirely on its own; it proposes actions (e.g., "Plan: Contact top 5 donors with this message...") which must be approved by a staff member before execution. All communications are logged for full auditability. This approach is also central to data privacy strategy, ensuring that automated processing is supervised and compliant.
The "Cold Start" Problem
The AI model needs data to be effective. To solve this, the system is designed to work from day one with a simple, rule-based logic (e.g., prioritize by last donation date and location). It runs the AI model in "shadow mode," learning from real interactions without impacting live operations until its predictions are proven to be accurate and reliable.
Log in or sign up for Devpost to join the conversation.